-
Notifications
You must be signed in to change notification settings - Fork 5
/
inject.c
760 lines (665 loc) · 28.3 KB
/
inject.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
#include "inject.h"
#include "read_dyld_info.h"
#include "headers/loader.h"
#include "headers/nlist.h"
#include "headers/reloc.h"
#include <stddef.h>
addr_t b_allocate_vmaddr(const struct binary *binary) {
addr_t max = 0;
for(uint32_t i = 0; i < binary->nsegments; i++) {
const range_t *range = &binary->segments[i].vm_range;
addr_t newmax = range->start + range->size;
if(newmax > max) max = newmax;
}
return (max + 0xfff) & ~0xfffu;
}
// this function is used by both b_macho_extend_cmds and b_inject_macho_binary
static void handle_retarded_dyld_info(void *ptr, uint32_t size, int num_segments, int *dylib_map, int num_dylib_map, bool kill_dones) {
// seriously, take a look at dyldinfo.cpp from ld64, especially, in this case, the separate handing of different LC_DYLD_INFO sections and the different meaning of BIND_OPCODE_DONE in lazy bind vs the other binds
// not to mention the impossibility of reading this data without knowing every single opcode
// and the lack of nop
void *end = ptr + size;
while(ptr != end) {
uint8_t byte = read_int(&ptr, end, uint8_t);
uint8_t immediate = byte & BIND_IMMEDIATE_MASK;
uint8_t opcode = byte & BIND_OPCODE_MASK;
switch(opcode){
// things we actually care about:
case BIND_OPCODE_DONE:
if(kill_dones) {
*((uint8_t *) ptr - 1) = BIND_OPCODE_SET_TYPE_IMM | BIND_TYPE_POINTER;
}
break;
case BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: {
// update the segment number
uint8_t *p = ptr - 1;
//printf("incr'ing %u by %u\n", (unsigned int) immediate, (unsigned int) num_segments);
*p = (*p & BIND_OPCODE_MASK) | (immediate + num_segments);
read_uleb128(&ptr, end);
break;
}
case BIND_OPCODE_SET_DYLIB_ORDINAL_IMM:
if(dylib_map) {
if(immediate == BIND_SPECIAL_DYLIB_SELF) break;
if(immediate >= num_dylib_map) {
die("invalid dylib reference");
}
if(dylib_map[immediate] > BIND_IMMEDIATE_MASK) {
//die("too many dylibs (imm)");
*((uint8_t *) ptr - 1) = BIND_OPCODE_SET_DYLIB_SPECIAL_IMM | ;
} else {
*((uint8_t *) ptr - 1) = opcode | dylib_map[immediate];
}
}
break;
case BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB:
if(dylib_map) {
uint8_t ordinal = read_int(&ptr, end, uint8_t);
if(ordinal == BIND_SPECIAL_DYLIB_SELF) break;
if(ordinal >= num_dylib_map) {
die("invalid dylib reference");
}
if(dylib_map[ordinal] > 0x7f) {
die("too many dylibs (uleb)");
}
*((uint8_t *) ptr - 1) = dylib_map[ordinal];
} else {
read_uleb128(&ptr, end);
}
break;
// things we have to get through
case BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM:
ptr += strnlen(ptr, end - ptr);
if(ptr == end)
break;
case BIND_OPCODE_SET_ADDEND_SLEB: // actually sleb (and I like how read_uleb128 and read_sleb128 in dyldinfo.cpp are completely separate functions), but read_uleb128 should work
case BIND_OPCODE_ADD_ADDR_ULEB:
case BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB:
read_uleb128(&ptr, end);
break;
case BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB:
read_uleb128(&ptr, end);
read_uleb128(&ptr, end);
break;
}
}
}
uint32_t b_macho_extend_cmds(struct binary *binary, size_t space) {
size_t old_size = b_mach_hdr(binary)->sizeofcmds;
size_t new_size = old_size + space;
if((new_size >> 12) == (old_size >> 12)) {
// good enough, it'll fit
return (new_size + 0xfff) & ~0xfff;
}
// looks like we need to make a duplicate header and do ugly stuff
size_t stuff_size = (sizeof(struct mach_header) + sizeof(struct segment_command) + new_size + 0xfff) & ~0xfff;
#define X(a) if(a) a += stuff_size;
CMD_ITERATE(b_mach_hdr(binary), cmd) {
switch(cmd->cmd) {
case LC_SEGMENT: {
struct segment_command *seg = (void *) cmd;
seg->fileoff += stuff_size;
struct section *sect = (void *) (seg + 1);
for(uint32_t i = 0; i < seg->nsects; i++, sect++) {
sect->offset += stuff_size;
X(sect->reloff)
}
break;
}
case LC_SYMTAB: {
struct symtab_command *sym = (void *) cmd;
X(sym->symoff)
X(sym->stroff)
break;
}
case LC_DYSYMTAB: {
struct dysymtab_command *dys = (void *) cmd;
X(dys->tocoff)
X(dys->modtaboff)
X(dys->extrefsymoff)
X(dys->indirectsymoff)
X(dys->extreloff)
X(dys->locreloff)
break;
}
case LC_TWOLEVEL_HINTS: {
struct twolevel_hints_command *two = (void *) cmd;
X(two->offset)
break;
}
case LC_CODE_SIGNATURE:
case LC_SEGMENT_SPLIT_INFO:
case 38 /*LC_FUNCTION_STARTS*/: {
// this is sort of a best (but rather bad) guess - all three commands will probably be screwed up by being moved like this
struct linkedit_data_command *dat = (void *) cmd;
X(dat->dataoff)
break;
}
case LC_ENCRYPTION_INFO: {
struct encryption_info_command *enc = (void *) cmd;
X(enc->cryptoff)
break;
}
case LC_DYLD_INFO:
case LC_DYLD_INFO_ONLY: {
struct dyld_info_command *dyl = (void *) cmd;
X(dyl->rebase_off)
X(dyl->export_off)
#define Y(a) if(dyl->a##_off) { \
prange_t pr = rangeconv_off((range_t) {binary, dyl->a##_off, dyl->a##_size}, MUST_FIND); \
handle_retarded_dyld_info(pr.start, pr.size, 1, NULL, 0, false); \
dyl->a##_off += stuff_size; \
}
Y(bind)
Y(weak_bind)
Y(lazy_bind)
#undef Y
break;
}
}
}
#undef X
binary->valid_range = pdup(binary->valid_range, ((binary->valid_range.size + 0xfff) & ~0xfff) + stuff_size, stuff_size);
struct mach_header *hdr = binary->valid_range.start;
struct segment_command *seg = (void *) (hdr + 1);
memcpy(hdr, binary->valid_range.start + stuff_size, sizeof(*hdr));
memcpy(seg + 1, binary->valid_range.start + stuff_size + sizeof(struct mach_header), hdr->sizeofcmds);
hdr->ncmds++;
hdr->sizeofcmds += sizeof(*seg);
seg->cmd = LC_SEGMENT;
seg->cmdsize = sizeof(*seg);
// yes, it MUST be called __TEXT.
static const char name[16] = "__TEXT";
memcpy(seg->segname, name, 16);
seg->vmaddr = b_allocate_vmaddr(binary);
seg->vmsize = stuff_size;
seg->fileoff = 0;
seg->filesize = stuff_size;
seg->maxprot = seg->initprot = PROT_READ | PROT_EXEC;
seg->nsects = 0;
seg->flags = 0;
return stuff_size - sizeof(struct mach_header);
}
// cctool's checkout.c insists on this exact order
enum {
MM_BIND, MM_WEAK_BIND, MM_LAZY_BIND,
MM_LOCREL,
MM_SYMTAB,
MM_LOCALSYM, MM_EXTDEFSYM, MM_UNDEFSYM,
MM_EXTREL,
MM_INDIRECT,
MM_STRTAB,
NMOVEME
};
struct linkedit_info {
arange_t linkedit_range;
void *linkedit_ptr;
// things we need to move:
// 0. string table
// 1-3. {local, extdef, undef}sym
// 4-5. {locrel, extrel}
// 6. indirect syms
// 7-9. dyld info {, weak_, lazy_}bind
// [hey, I will just assume that nobody has any section relocations because it makes things simpler!]
// things we need to update:
// - symbols reference string table
// - relocations reference symbols
// - indirect syms reference symbols
// - (section data references indirect syms)
struct moveme {
uint32_t *off, *size;
uint32_t element_size;
int off_base;
void *copied_to;
void *copied_from;
uint32_t copied_size;
} moveme[NMOVEME];
struct symtab_command *symtab;
struct dysymtab_command *dysymtab;
struct dyld_info_command *dyld_info;
};
static const struct moveref {
int target_start, target_end;
ptrdiff_t offset;
} moveref[NMOVEME] = {
[MM_LOCALSYM] = {MM_STRTAB, MM_STRTAB, offsetof(struct nlist, n_un.n_strx)},
[MM_EXTDEFSYM] = {MM_STRTAB, MM_STRTAB, offsetof(struct nlist, n_un.n_strx)},
[MM_UNDEFSYM] = {MM_STRTAB, MM_STRTAB, offsetof(struct nlist, n_un.n_strx)},
// hooray for little endian
[MM_LOCREL] = {MM_LOCALSYM, MM_UNDEFSYM, 4},
[MM_EXTREL] = {MM_LOCALSYM, MM_UNDEFSYM, 4},
// the whole thing is a symbol number
[MM_INDIRECT] = {MM_LOCALSYM, MM_UNDEFSYM, 0}
};
static bool catch_linkedit(struct mach_header *hdr, struct linkedit_info *li, bool patch) {
memset(li, 0, sizeof(*li));
bool ret = false;
CMD_ITERATE(hdr, cmd) {
restart:
switch(cmd->cmd) {
case LC_SEGMENT: {
struct segment_command *seg = (void *) cmd;
if(!strcmp(seg->segname, "__LINKEDIT")) {
li->linkedit_range.start = seg->fileoff;
li->linkedit_range.size = seg->filesize;
ret = true;
goto patchout;
break;
}
break;
}
case LC_SYMTAB: {
struct symtab_command *symtab = (void *) cmd;
li->symtab = symtab;
li->moveme[MM_STRTAB].off = &symtab->stroff;
li->moveme[MM_STRTAB].size = &symtab->strsize;
li->moveme[MM_STRTAB].element_size = 1;
li->moveme[MM_SYMTAB].off = &symtab->symoff;
li->moveme[MM_SYMTAB].size = &symtab->nsyms;
li->moveme[MM_SYMTAB].element_size = sizeof(struct nlist);
li->moveme[MM_SYMTAB].off_base = -1;
break;
}
case LC_DYSYMTAB: {
struct dysymtab_command *dys = (void *) cmd;
li->dysymtab = dys;
li->moveme[MM_LOCALSYM].off = &dys->ilocalsym;
li->moveme[MM_LOCALSYM].size = &dys->nlocalsym;
li->moveme[MM_LOCALSYM].element_size = sizeof(struct nlist);
li->moveme[MM_LOCALSYM].off_base = MM_SYMTAB;
li->moveme[MM_EXTDEFSYM].off = &dys->iextdefsym;
li->moveme[MM_EXTDEFSYM].size = &dys->nextdefsym;
li->moveme[MM_EXTDEFSYM].element_size = sizeof(struct nlist);
li->moveme[MM_EXTDEFSYM].off_base = MM_SYMTAB;
li->moveme[MM_UNDEFSYM].off = &dys->iundefsym;
li->moveme[MM_UNDEFSYM].size = &dys->nundefsym;
li->moveme[MM_UNDEFSYM].element_size = sizeof(struct nlist);
li->moveme[MM_UNDEFSYM].off_base = MM_SYMTAB;
li->moveme[MM_LOCREL].off = &dys->locreloff;
li->moveme[MM_LOCREL].size = &dys->nlocrel;
li->moveme[MM_LOCREL].element_size = sizeof(struct relocation_info);
li->moveme[MM_EXTREL].off = &dys->extreloff;
li->moveme[MM_EXTREL].size = &dys->nextrel;
li->moveme[MM_EXTREL].element_size = sizeof(struct relocation_info);
li->moveme[MM_INDIRECT].off = &dys->indirectsymoff;
li->moveme[MM_INDIRECT].size = &dys->nindirectsyms;
li->moveme[MM_INDIRECT].element_size = 4;
break;
}
case LC_DYLD_INFO_ONLY:
case LC_DYLD_INFO: {
struct dyld_info_command *di = (void *) cmd;
li->dyld_info = di;
if(patch) {
di->rebase_off = 0;
di->rebase_size = 0;
di->export_off = 0;
di->export_size = 0;
}
li->moveme[MM_BIND].off = &di->bind_off;
li->moveme[MM_BIND].size = &di->bind_size;
li->moveme[MM_BIND].element_size = 1;
li->moveme[MM_WEAK_BIND].off = &di->weak_bind_off;
li->moveme[MM_WEAK_BIND].size = &di->weak_bind_size;
li->moveme[MM_WEAK_BIND].element_size = 1;
li->moveme[MM_LAZY_BIND].off = &di->lazy_bind_off;
li->moveme[MM_LAZY_BIND].size = &di->lazy_bind_size;
li->moveme[MM_LAZY_BIND].element_size = 1;
break;
}
patchout:
case LC_CODE_SIGNATURE:
case LC_SEGMENT_SPLIT_INFO:
case 38 /*LC_FUNCTION_STARTS*/:
// hope you didn't need that stuff <3
if(patch) {
hdr->sizeofcmds -= cmd->cmdsize;
size_t copysize = hdr->sizeofcmds - ((char *) cmd - (char *) (hdr + 1));
hdr->ncmds--;
memcpy(cmd, (char *) cmd + cmd->cmdsize, copysize);
// don't run off the end
if(!copysize) goto end;
goto restart;
}
break;
}
}
end:
// we want both binaries to have a symtab and dysymtab, makes things easier
if(!li->symtab || !li->dysymtab) die("symtab/dysymtab missing");
return ret;
}
static void fixup_stub_helpers(int cputype, void *base, size_t size, uint32_t incr) {
if(!size) return;
size_t skip_begin, skip_end, offset, stride;
switch(cputype) {
case CPU_TYPE_ARM:
skip_begin = 0x24;
skip_end = 0;
offset = 8;
stride = 0xc;
break;
case CPU_TYPE_X86:
skip_begin = 0;
skip_end = 0xa;
offset = 1;
stride = 0xa;
break;
default:
die("stub_helpers, but unknown cpu type");
}
if(size < (skip_begin + skip_end)) {
die("unknown stub_helpers format (too small)");
}
base += skip_begin; size -= skip_begin;
while(size >= skip_end + stride) {
*((uint32_t *) (base + offset)) += incr;
base += stride; size -= stride;
}
}
void b_inject_macho_binary(struct binary *target, const struct binary *binary, addr_t (*find_hack_func)(const struct binary *binary), bool userland) {
#define ADD_COMMAND(size) ({ \
void *ret = (char *) hdr + sizeof(struct mach_header) + hdr->sizeofcmds; \
uint32_t newsize = hdr->sizeofcmds + size; \
if(newsize > sizeofcmds_limit) { \
die("not enough space for commands"); \
} \
hdr->ncmds++; \
hdr->sizeofcmds += (uint32_t) (size); \
ret; \
})
#define ADD_SEGMENT(size) ({ \
uint32_t ret = (seg_off + 0xfff) & ~0xfff; \
seg_off = ret + (size); \
ret; \
})
#define ADD_SEGMENT_ADDR(size) ({ \
uint32_t ret = (seg_addr + 0xfff) & ~0xfff; \
seg_addr = ret + (size); \
ret; \
})
uint32_t sizeofcmds_limit = b_macho_extend_cmds(target, b_mach_hdr(binary)->sizeofcmds);
size_t seg_off = target->valid_range.size;
addr_t seg_addr = 0;
struct mach_header *hdr = b_mach_hdr(target);
hdr->flags &= ~MH_PIE;
const struct binary *binaries[] = {binary, target};
// in userland mode, we cut off the LINKEDIT segment (for target, only if it's at the end of the binary)
struct linkedit_info li[2];
if(userland) {
for(int i = 0; i < 2; i++) {
if(catch_linkedit(b_mach_hdr(binaries[i]), &li[i], i == 1)) {
li[i].linkedit_ptr = rangeconv_off((range_t) {binaries[i], li[i].linkedit_range.start, li[i].linkedit_range.size}, MUST_FIND).start;
}
}
if((size_t) (li[1].linkedit_range.start + li[1].linkedit_range.size) == seg_off) {
target->valid_range.size = seg_off = li[1].linkedit_range.start;
}
if((li[0].dyld_info != 0) != (li[1].dyld_info != 0)) {
die("LC_DYLD_INFO(_ONLY) should be in both or neither");
}
}
uint32_t init_ptrs[100];
unsigned num_init_ptrs = 0;
uint32_t *reserved1s[100];
unsigned num_reserved1s = 0;
struct copy { ptrdiff_t off; void *start; size_t size; } copies[100];
unsigned num_copies = 0;
// the names of the source's loaded dylibs
const char *dylib_names[100];
unsigned num_dylibs = 0;
// dylib ordinal from target -> source
int dylib_map[100] = {0};
unsigned num_dylib_map = 1; // size of dylib_map
unsigned num_segments = 0;
if(userland) {
CMD_ITERATE(hdr, cmd) {
switch(cmd->cmd) {
case LC_SEGMENT: {
num_segments++;
struct segment_command *seg = (void *) cmd;
struct section *sections = (void *) (seg + 1);
for(uint32_t i = 0; i < seg->nsects; i++) {
struct section *sect = §ions[i];
switch(sect->flags & SECTION_TYPE) {
case S_NON_LAZY_SYMBOL_POINTERS:
case S_LAZY_SYMBOL_POINTERS:
case S_SYMBOL_STUBS:
if(num_reserved1s < 100) reserved1s[num_reserved1s++] = §->reserved1;
break;
}
// xxx - what happens if there is no dyld_info?
if(li[0].dyld_info && !strcmp(sect->sectname, "__stub_helper")) {
void *segdata = rangeconv_off((range_t) {target, seg->fileoff, seg->filesize}, MUST_FIND).start;
fixup_stub_helpers(hdr->cputype, segdata + sect->offset - seg->fileoff, sect->size, *li[0].moveme[MM_LAZY_BIND].size);
}
}
break;
}
case LC_LOAD_DYLIB:
if(num_dylibs < 100) dylib_names[num_dylibs++] = convert_lc_str(cmd, ((struct dylib_command *) cmd)->dylib.name.offset);
break;
}
}
}
CMD_ITERATE(b_mach_hdr(binary), cmd) {
switch(cmd->cmd) {
case LC_SEGMENT: {
struct segment_command *seg = (void *) cmd;
if(userland && !strcmp(seg->segname, "__LINKEDIT")) continue;
size_t size = sizeof(struct segment_command) + seg->nsects * sizeof(struct section);
// make seg_addr useful
addr_t new_addr = seg->vmaddr + seg->vmsize;
if(new_addr > seg_addr) seg_addr = new_addr;
struct segment_command *newseg = ADD_COMMAND(size);
memcpy(newseg, seg, size);
prange_t pr = rangeconv_off((range_t) {binary, seg->fileoff, seg->filesize}, MUST_FIND);
newseg->fileoff = (uint32_t) ADD_SEGMENT(pr.size);
//printf("setting fileoff to %u\n", newseg->fileoff);
if(num_copies < 100) copies[num_copies++] = (struct copy) {newseg->fileoff, pr.start, pr.size};
struct section *sections = (void *) (newseg + 1);
for(uint32_t i = 0; i < seg->nsects; i++) {
struct section *sect = §ions[i];
sect->offset = newseg->fileoff + sect->addr - newseg->vmaddr;
// ZEROFILL is okay because iBoot always zeroes vmsize - filesize
if(!userland && (sect->flags & SECTION_TYPE) == S_MOD_INIT_FUNC_POINTERS) {
uint32_t *p = rangeconv_off((range_t) {binary, sect->offset, sect->size}, MUST_FIND).start;
size_t num = sect->size / 4;
while(num--) {
if(num_init_ptrs < 100) init_ptrs[num_init_ptrs++] = *p++;
}
}
}
break;
}
case LC_LOAD_DYLIB:
if(userland) {
unsigned int targetlib;
const char *name = convert_lc_str(cmd, ((struct dylib_command *) cmd)->dylib.name.offset);
for(targetlib = 0; targetlib < num_dylibs; targetlib++) {
if(!strcmp(name, dylib_names[targetlib])) {
goto ok;
}
}
// otherwise, we have to add a new LC
void *newcmd = ADD_COMMAND(cmd->cmdsize);
memcpy(newcmd, cmd, cmd->cmdsize);
num_dylibs++;
ok:
if(num_dylib_map < 100) dylib_map[num_dylib_map++] = targetlib + 1;
}
break;
}
}
// now deal with the init pointers (if not userland)
// this code is really gross
if(num_init_ptrs > 0) {
if(num_init_ptrs == 1) { // hey, correct plurals are nice
fprintf(stderr, "note: 1 constructor function is present; using the hack_func\n");
} else {
fprintf(stderr, "note: %d constructor functions are present; using the hack_func\n", num_init_ptrs);
}
if(!find_hack_func) {
die("...but there was no find_hack_func");
}
// ldr pc, [pc]
uint16_t part0[] = {0xf8df, 0xf000};
// push {r0-r3, lr}; adr lr, f+1; ldr pc, a; f: b next; a: .long 0; next:
// (the address of the init func)
//
uint16_t part1[] = {0xb50f, 0xf20f, 0x0e07, 0xf8df, 0xf004, 0xe001};
// (bytes_to_move bytes of stuff)
// pop {r0-r3, lr}
static const uint16_t part2[] = {0xe8bd, 0x400f};
// ldr pc, [pc]
static const uint16_t part3[] = {0xf8df, 0xf000};
uint32_t bytes_to_move = 12; // don't cut the MRC in two!
addr_t hack_func = find_hack_func(target);
fprintf(stderr, "hack_func = %08x\n", hack_func);
prange_t hack_func_pr = rangeconv((range_t) {target, hack_func & ~1, bytes_to_move}, MUST_FIND);
// allocate a new segment for the stub
uint32_t stub_size = (uint32_t) ((sizeof(part1) + 4) * num_init_ptrs + sizeof(part2) + bytes_to_move + sizeof(part3) + 4);
if(!(hack_func & 1)) {
die("hack func 0x%x is not thumb", hack_func);
}
struct segment_command *newseg = ADD_COMMAND(sizeof(struct segment_command));
newseg->cmd = LC_SEGMENT;
newseg->cmdsize = sizeof(struct segment_command);
memset(newseg->segname, 0, 16);
strcpy(newseg->segname, "__CRAP");
newseg->vmaddr = ADD_SEGMENT_ADDR(stub_size);
newseg->vmsize = stub_size;
newseg->fileoff = ADD_SEGMENT(stub_size);
newseg->filesize = stub_size;
newseg->maxprot = newseg->initprot = PROT_READ | PROT_EXEC;
newseg->nsects = 0;
newseg->flags = 0;
void *ptr = malloc(stub_size);
for(unsigned i = 0; i < num_init_ptrs; i++) {
memcpy(ptr, part1, sizeof(part1));
ptr += sizeof(part1);
memcpy(ptr, &init_ptrs[i], 4);
ptr += 4;
part1[0] = 0x46c0;
}
memcpy(ptr, part2, sizeof(part2));
ptr += sizeof(part2);
memcpy(ptr, hack_func_pr.start, bytes_to_move);
ptr += bytes_to_move;
memcpy(ptr, part3, sizeof(part3));
ptr += sizeof(part3);
uint32_t new_addr = hack_func + bytes_to_move;
memcpy(ptr, &new_addr, 4);
ptr += 4;
new_addr = newseg->vmaddr | 1;
memcpy(hack_func_pr.start, part0, sizeof(part0));
memcpy(hack_func_pr.start + sizeof(part0), &new_addr, 4);
if(num_copies < 100) copies[num_copies++] = (struct copy) {newseg->fileoff, ptr, stub_size};
}
autofree char *linkedit = NULL;
if(userland) {
// build the new LINKEDIT
uint32_t newsize = 0;
for(int i = 0; i < NMOVEME; i++) {
for(int l = 0; l < 2; l++) {
if(li[l].moveme[i].off_base != -1) {
newsize += *li[l].moveme[i].size * li[l].moveme[i].element_size;
}
}
}
if(newsize != 0) {
uint32_t linkedit_off = ADD_SEGMENT(newsize);
linkedit = malloc(newsize);
uint32_t off = 0;
for(int i = 0; i < NMOVEME; i++) {
uint32_t s = 0;
for(int l = 0; l < 2; l++) {
struct moveme *m = &li[l].moveme[i];
m->copied_size = *m->size * m->element_size;
m->copied_to = linkedit + off + s;
if(m->off_base > 0) {
// the value is an index into a table represented by another moveme (i.e. the symtab)
m->copied_from = li[l].moveme[m->off_base].copied_from + *m->off * m->element_size;
} else {
// the value is a file offset
m->copied_from = li[l].linkedit_ptr - li[l].linkedit_range.start + *m->off;
}
if(m->off_base != -1) memcpy(m->copied_to, m->copied_from, m->copied_size);
s += m->copied_size;
}
//printf("i=%d s=%u off=%u\n", i, s, off);
// update the one to load
struct moveme *m = &li[1].moveme[i];
*m->off = linkedit_off + off;
if(m->off_base > 0) {
*m->off = (*m->off - *li[1].moveme[m->off_base].off) / m->element_size;
}
*m->size = s / m->element_size;
if(m->off_base != -1) off += s;
}
// update struct references (which are out of order, yay)
off = 0;
for(int i = MM_LOCREL; i <= MM_INDIRECT; i++) {
if(moveref[i].target_start) {
struct moveme *restrict m = &li[1].moveme[i];
for(void *ptr = m->copied_to; ptr < m->copied_to + m->copied_size; ptr += m->element_size) {
uint32_t diff = 0;
for(int j = moveref[i].target_start; j <= moveref[i].target_end; j++) {
diff += *li[0].moveme[j].size;
}
uint32_t *p = ptr + moveref[i].offset;
if(*p < 0x10000000) *p += diff;
}
}
}
// update library numbers in symbol table
{
struct moveme *restrict m = &li[0].moveme[MM_UNDEFSYM];
for(struct nlist *nl = m->copied_to; (void *) (nl + 1) <= (m->copied_to + m->copied_size); nl++) {
unsigned lib = GET_LIBRARY_ORDINAL(nl->n_desc);
if(lib != SELF_LIBRARY_ORDINAL && lib <= MAX_LIBRARY_ORDINAL) {
if(lib >= num_dylib_map) {
die("invalid dylib reference");
}
unsigned new = dylib_map[lib];
if(new > MAX_LIBRARY_ORDINAL) {
die("too many libraries");
}
SET_LIBRARY_ORDINAL(nl->n_desc, new + 1);
}
}
}
// ... and update section references
for(unsigned i = 0; i < num_reserved1s; i++) {
*reserved1s[i] += *li[0].moveme[MM_INDIRECT].size;
}
// ... and dyld info
if(li->dyld_info) {
for(int i = MM_BIND; i <= MM_LAZY_BIND; i++) {
if(*li[1].moveme[i].off) {
handle_retarded_dyld_info(linkedit - linkedit_off + *li[1].moveme[i].off, *li[0].moveme[i].size, num_segments, dylib_map, num_dylib_map, i != MM_LAZY_BIND);
}
}
}
struct segment_command *newseg = ADD_COMMAND(sizeof(struct segment_command));
newseg->cmd = LC_SEGMENT;
newseg->cmdsize = sizeof(struct segment_command);
memset(newseg->segname, 0, 16);
strcpy(newseg->segname, "__LINKEDIT");
newseg->vmaddr = ADD_SEGMENT_ADDR(newsize);
newseg->vmsize = (newsize + 0xfff) & ~0xfff;
newseg->fileoff = linkedit_off;
newseg->filesize = newsize;
newseg->maxprot = newseg->initprot = PROT_READ | PROT_WRITE;
newseg->nsects = 0;
newseg->flags = 0;
//printf("off=%d newsize=%d\n", linkedit_off, newsize);
if(num_copies < 100) copies[num_copies++] = (struct copy) {linkedit_off, linkedit, newsize};
}
}
// finally, expand the binary in memory and actually copy in the new stuff
target->valid_range = pdup(target->valid_range, seg_off, 0);
for(unsigned i = 0; i < num_copies; i++) {
memcpy(target->valid_range.start + copies[i].off, copies[i].start, copies[i].size);
}
}